Aggregated Modeling And Control Of Thermostatically Controlled Loads

As a main type of flexible load, a large number of thermostatically controlled loads (TCLs) have the ability to provide ancillary services such as frequency control and power balance. They can also achieve peak load shifting as well as reduce wear and tear of the generation equipment caused by frequently start-stop actions. Moreover, power fluctua-tions caused by renewable energy can be restrained by TCLs too. Realization of these functions depends on a proper control strategy and an accurate load model. Practical performance of the control strategy based on an inaccurate load model may fade in the aspect of application. Therefore, aggregated modeling and control of TCLs are studied in this paper, and the main research results are listed as follows.Firstly, a multi-input single-output (MISO) bilinear aggregate model is built for a fam-ily of homogeneous TCLs. By adding a bilinear control item to the traditional single-input single-output (SISO) bilinear aggregate model, the MISO model eliminate the shortcom-ing of the SISO model that the accuracy decreases with the increase of simulation time. Besides, in the situation that the setpoint temperature is changing, the proposed model not only considers impacts of real-time variation of setpoint temperature on aggregation behaviors, but also takes into account the effect of cumulative variation of setpoint temper-ature, which improves the performance of the aggregate model considerably. Simulation results show that the proposed bilinear aggregate model has a wide applicability, which can guarantee a high accuracy regardless of the type of the setpoit temperature curve such as linear-line, polyline or other arbitrary curves.Secondly, a modeling method based on homogeneous aggregate models is designed for heterogeneous TCLs. According to the correlation of heterogeneous TCLs with homoge-neous TCLs, a large number of heterogeneous TCLs can be approximately decomposed into multiple clusters of homogeneous loads utilizing the method of clustering. This modeling method provides a basic framework for modeling heterogeneous TCLs, which can derive different forms of heterogeneous aggregate models combining with different homogeneous aggregate models. Numerical simulations are provided to show that the proposed method can not only be applied to model heterogeneous TCLs of the same type, but also be suitable for heterogeneous TCLs of multiple types.Finally, a distributed control algorithm with agents in two levels, load agents in the primary level and load aggregators in the secondary level, is planned to solve the power tracking problem of a family of TCLs. First of all. based on the day-ahead generation plan of generators, load forecasting of load agents and the cost quotation of generation units, a real-time reference power trajectory for generators and load agents are provided by dispatching center using centralized optimal economic dispatch. Then, a distributed pinning coordinated algorithm is utilized by the up-level load agent and low-level load aggregators to determine control signals for each aggregator. Finally, in each aggregator (e.g. intelligent community), every TCL is controlled in a centralized way without com-municating with the neighbor. Due to the characteristic of the distributed structure, the proposed power tracking scheme is stable, scalable, robust, low-cost and reliable. This algorithm is simulated on a IEEE 9 bus system, the simulation result shows that by adopt-ing the distributed control strategy, load agents can track the reference power trajectory successfully.